BUG: read_excel for ods files raising UnboundLocalError in certain cases

pandas/io/excel/_odfreader.py

@@ -28,8 +28,28 @@ def __init__(
         storage_options: StorageOptions = None,
     ):
         import_optional_dependency("odf")
+        self._monkeypatch_odf_element_str()
         super().__init__(filepath_or_buffer, storage_options=storage_options)
 
+    def _monkeypatch_odf_element_str(self):
+        from odf.element import Element
+        from odf.namespaces import TEXTNS
+        from odf.text import S
+
+        text_s = S().qname
+
+        def element_str(self):
+            value = []
+            for c in self.childNodes:
+                if isinstance(c, Element) and c.qname == text_s:
+                    spaces = int(c.attributes.get((TEXTNS, "c"), 1))
+                    value.append(" " * spaces)
+                else:
+                    value.append(str(c))
+            return "".join(value)
+
+        Element.__str__ = element_str
+
     @property
     def _workbook_class(self):
         from odf.opendocument import OpenDocument
@@ -178,7 +198,7 @@ def _get_cell_value(self, cell, convert_float: bool) -> Scalar:
             cell_value = cell.attributes.get((OFFICENS, "value"))
             return float(cell_value)
         elif cell_type == "string":
-            return self._get_cell_string_value(cell)
+            return str(cell)
         elif cell_type == "currency":
             cell_value = cell.attributes.get((OFFICENS, "value"))
             return float(cell_value)
@@ -191,28 +211,3 @@ def _get_cell_value(self, cell, convert_float: bool) -> Scalar:
             return result.time()
         else:
             raise ValueError(f"Unrecognized type {cell_type}")
-
-    def _get_cell_string_value(self, cell) -> str:
-        """
-        Find and decode OpenDocument text:s tags that represent
-        a run length encoded sequence of space characters.
-        """
-        from odf.element import Element, Text
-        from odf.namespaces import TEXTNS
-        from odf.text import P, S
-
-        text_p = P().qname
-        text_s = S().qname
-
-        p = cell.childNodes[0]
-
-        value = []
-        if p.qname == text_p:
-            for k, fragment in enumerate(p.childNodes):
-                if isinstance(fragment, Text):
-                    value.append(fragment.data)
-                elif isinstance(fragment, Element):
-                    if fragment.qname == text_s:
-                        spaces = int(fragment.attributes.get((TEXTNS, "c"), 1))
-                    value.append(" " * spaces)
-        return "".join(value)

pandas/tests/io/excel/test_readers.py

@@ -499,6 +499,23 @@ def test_reader_spaces(self, read_ext):
         )
         tm.assert_frame_equal(actual, expected)
 
+    # gh-36122, gh-35802
+    @pytest.mark.parametrize(
+        "basename,expected",
+        [
+            ("gh-35802", DataFrame({"COLUMN": ["Test (1)"]})),
+            ("gh-36122", DataFrame(columns=["got 2nd sa"])),
+        ],
+    )
+    def test_read_excel_ods_nested_xml(self, read_ext, basename, expected):
+        # see gh-35802
+        engine = pd.read_excel.keywords["engine"]
+        if engine != "odf":
+            pytest.skip(f"Skipped for engine: {engine}")
+
+        actual = pd.read_excel(basename + read_ext)
+        tm.assert_frame_equal(actual, expected)
+
     def test_reading_all_sheets(self, read_ext):
         # Test reading all sheet names by setting sheet_name to None,
         # Ensure a dict is returned.